MT7621 - 21.02/Master feedback firmware image test - IPV6 offload and disabled Flow Control


List your packages installed?
ISP is PPPoE or IPoE?

Probably have regression. Or incompatiblite with tunnels ipv6 in ipv4 or vice versa.
Or it's on nftables. Remenber that on master is runnig migration of iptables to nftables.

I not detect this crash on my router with master.

like i said above i'm not using PPPoE. my ipv6 wan is DHCPv6-PD.

i just have a basic setup with 2 lans using vlan filtering.

for me, latest patches to mt7621 offload and switch changes on ubiquity erx brought instability with hw offloading

Putting my good experience in context, I've only briefly tested hw offload and then switched right back to sw offload only. I am not using PPPoE either.

That said, I've got 12 days of ER-X gateway runtime without issues on r18993-a1b8a4d7b3 using only sw offload.

I'm using HW offlload (IPv4 only) with a build from Feb 18th for 19 days now!

I have selected this build as my "stable" build (better than 21.02.2 for my needs).

1 Like

the issues seems to be with ipv6 combo, there was a previous patch for ipv6 but they seems to have added a different one, I think is not well tested for ipv4+ipv6 scenario in HW offload (previous I had uptime of 40+ days, until a new flash)

this new commit improve perfomance.
I not have time for test now. Only for report for team;a=commit;h=fdbb468c5998d9adb1819751ffaf38013768a8eb


Does it have any effect on mt7621 series?


"This can improve performance under load by ensuring that NAPI processing is not pinned on CPU 0.";a=blob;f=target/linux/generic/pending-5.10/702-net-ethernet-mtk_eth_soc-enable-threaded-NAPI.patch;h=da9c040322a24c8d4387df860df44f243bebca4a;hb=fdbb468c5998d9adb1819751ffaf38013768a8eb

I did not see any improvement on MT7621. See here.

Thanks for report.

1 Like

I have a D-Link DIR-1960 (same as 2660) with the mt7621. Is there a snapshot available for my device with the more recent additions/changes?

IPV6 offload support is now in the official openwrt repos and latest images

Disabled flow control is only on my repo -

You need to build the code to make an image.

1 Like

Hey @db260179 there was a thread listed regarding 5.10 kernel builds, and someone apparently back-ported 5.12 code to fix offload on it. Not sure how this impacts your builds.

My original patched did have that, but when pushing upstream, was told

Please remove the FLOW_ACT_MANGLE_HDR_TYPE_IP6 related code. The hardware does not support IPv6 NAT, only routing.

So not sure what is going on there?

Hello, need a little help.

I had a hootoo/strong 1200 device running openwrt.

I tried to create a patch for my device, edited the dts files, added the compatiblity for network etc.

After building i thought of flashing the initramfs image in order to "test" the firmware before running sysupgrade, thinking initramfs doesn't "write" anything on the device but i was wrong.

Unfortunately flashing the initramfs kernel on the device bricked it and i cannot connect to it via putty.

Does flashing initramfs image via openwrt can brick the device?

Any tips on recovering the device? do I need to connect to serial port?


RAM booting an initramfs image over tftp doesn't write to flash by itself (although you can manually write to flash from the running initramfs booted system), but that's not what you have done.

So you did write the initramfs image to flash, before having tested it - that replaced the installed image and you're left with whatever the initramfs image is giving you.

You should have only RAM booted it from the device's bootloader via tftp over the network, that way the initramfs image is only executed in RAM and never written to flash.

Thanks for your reply. so hard way to learn that that flashing initramfs image can also brick the device.

I will try to solder headers into the device and attempt recovery. Hope I am able to get the device back running.

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.